OpenGradient

OpenGradient is a decentralized network for verifiable AI. Its homepage says it hosts models, runs secure inference, and deploys agents onchain, and it describes the network as 100% EVM compatible. As of 2026-10-29 the site shows 4,500 models, more than 2 million verifiable AI inferences, and more than 500K zkML proofs and TEE attestations. The older documentation overview still quotes 2,000+ models, so treat the homepage figure as the current one.

Key Features

• Model Hub: a permissionless repository for models of any architecture, hosted onchain and available for inference on the network.
• On-Chain AI SDK: a Python SDK for running ML and LLM inference, managing models, and deploying automated workflows with integrity guarantees.
• Verifiable execution: inference results can be backed by zkML proofs or TEE attestations, which is the core of the network's pitch.
• Confidential LLM inference: per the docs, the TEE path lets you use the OpenAI Python SDK with prompts encrypted to an attested TEE and relayed through the OpenGradient Chat API. The docs also list x402 under secure LLM inference.
• Ecosystem products: the homepage lists DreamFort (private chat infrastructure), Digital Twins at twin.fun (AI replicas of influencers), and BitQuant (a personal AI quantitative analyst). The docs also mention MemSync, a long-term memory layer for AI apps.

Use Cases

• Agents and apps that need a proof or attestation that a specific model produced a specific output.
• Teams that want to host or discover models onchain rather than behind a closed API.
• Developers who want an OpenAI-SDK-style workflow with encrypted prompts.

Limitation: the homepage and docs do not publish pricing or token details, so budget and rate questions have to be answered from the network's own pages before you build on it.

Pricing

No price list appears on the pages checked on 2026-10-29. Confirm costs with OpenGradient directly.

Getting Started

1. Read the docs overview and pick the SDK or the Chat API route.
2. Install the Python SDK from the OpenGradient GitHub organization.
3. Run a small inference on a Model Hub model and inspect the proof or attestation before wiring it into a product.

FAQ

What is OpenGradient used for?

Running and hosting AI models and agents onchain with verifiable inference, using zkML proofs or TEE attestations.

How many models does OpenGradient host?

The homepage shows 4,500 models as of 2026-10-29. The docs overview still says 2,000+.

Does OpenGradient publish prices?

Not on the homepage or docs overview we checked, so this page lists none.

Alternatives

• OpenRouter: a multi-model API without onchain proofs.
• Langfuse: log-based LLM observability.
• ToAPIs: an OpenAI-compatible gateway.

Conclusion

OpenGradient fits builders who need verifiable AI inference on an EVM network. Start from the official site, and compare simpler gateways in the AI ecosystem category if you do not need proofs.